Netflix - Information Technology

1

Database Marketing

Chapter Extension 12

ce12-2

Study Questions

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

Q1: What is a database marketing opportunity?

Q2: How does RFM analysis classify customers?

Q3: How does market-basket analysis identify cross-selling opportunities?

Q4: How do decision trees identify market segments?

2

ce12-3

Owner of Carbon Creek Gardens Needs Database Marketing

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• As business grew, lost track of customers

• Lost valuable customer and didn't know it

• Has lot of sales data, but needs system to: – Store and track customers – Store and track services provided to customers – Store and report future scheduled services

ce12-4

Database Marketing

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Application of business intelligence systems to: – Planning marketing programs – Executing marketing programs – Assessing marketing programs

• Databases and data mining techniques key components

3

ce12-5

Q2: How Does RFM Analysis Classify Customers?

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Recently

• Frequently

• Money

ce12-6

RFM Analysis Classifies Customers

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

4

ce12-7

Q3: How Does Market-Basket Analysis Identify Cross-Selling Opportunities?

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Unsupervised data mining method – Statistical methods to identify sales patterns in large volumes

of data – Products customers tend to buy together – Probabilities of customer purchases – Identify cross-selling opportunities

How many customers bought both fins and a mask?

ce12-8

Market-Basket Example: Transactions = 400

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

5

ce12-9

Support: Probability that Two Items Will Be Bought Together

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• P(Fins and Masks) = 250/400, or 62% • P(Fins and Weights) = 20/400, or 5%

ce12-10

Confidence = Conditional Probability Estimate

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Probability of buying Fins = 250 -- Probability of buying Mask = 270 • P(After buying Mask, then will buy Fins) -- Confidence = 250/270 or 92.6%

6

ce12-11

Lift = Confidence ÷ Base Probability

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Lift = Confidence of Mask/Base P(Fins)

• Lift = .926/.625 = 1.32

ce12-12

Warning!

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Analysis only shows shopping carts with two items

• Must analyze large number of shopping carts with three or more items

• Know what problem you are solving before mining the data – Know what question you want to answer

7

ce12-13

Q4: How Do Decision Trees Identify Market Segments?

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Hierarchical arrangement of criteria to predict a classification or value

• If/Then hierarchy

• Unsupervised data mining technique

• Basic idea of a decision tree – Select attributes most useful for classifying something on

some criterion to create “pure groups”

ce12-14

Decision Tree for Student

Performance

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

If Junior = Yes

Lower-level groups more similar than higher-level groups

8

ce12-15

Transforming a Set of Decision Rules

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• If student is a junior and works in a restaurant, – Then predict grade =>3.0

• If student is a senior and is a nonbusiness major, – Then predict grade <3.0

• If student is a junior and does not work in a restaurant, – Then predict grade <3.0

• If student is a senior and is a business major, – Then make no prediction

ce12-16

Decision Tree for Loan Evaluation

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• Classify loan applications by likelihood of default • Rules identify loans for bank approval • Identify market segment • Structure marketing campaign • Predict problems

9

ce12-17

Credit Score Decision Tree

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

ce12-18

Ethics Guide: Data Mining in the Real World

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

Problems with data: • Dirty data • Missing values • Lack of knowledge at start of project • Overfitting – too many variables • Probabilistic—good model may have unlucky first uses • Seasonality influences • High risk – uncovering something self-defeating to reveal

10

ce12-19

Ethics Guide: Data Mining in the Real World (cont’d)

C o p y r i g h t © 2 0 1 7 P e a r s o n E d u c a t i o n , I n c .

• When you start a data mining project, you never know how it will turn out

• Decision trees can be used to select variables for other types of data mining analysis